SSP Smaato adds machine learning for directing inventory to DSPs
With this boost, the company says, its Automated Traffic Curation can better match in-app ad requests with the right Demand Side Platform.
Supply-side platform Smaato announced Tuesday the addition of machine learning to its Automated Traffic Curation (ATC), a step that it says can “significantly” reduce the traffic overload for Demand Side Platforms (DSPs) and result in better pricing matches for publishers.
What this means. For the past year or so, the San Francisco-based firm has been employing its ATC to better manage the ad requests from the inventory for the 90,000+ app publishers it represents.
Let’s say a placement in an app from Publisher A calls for an ad in the Smaato exchange, and the exchange then sends the request to any of the 260+ DSPs in its network.
But, Chief Product Officer Freddy Friedman told me, DSPs are getting swamped with ad requests from apps and other inventory, because of the increased traffic generated by header bidding and other factors.
Some DSPs are undertaking their own efforts to “throttle” the flood by arbitrarily filtering out some ad requests. This allows those DSPs to better handle the bids coming in for the remaining inventory ad requests, but it also means that some inventory is pushed aside.
Why machine learning helps the process. Before today, Smaato’s ATC helped to steer ad requests to selected DSPs instead of broadcasting them out across the board, using manually-set rules about CPM rates, clickthrough rates, territory, kinds of ads, time of day, and other characteristics of each DSP. The same ad request is often sent to multiple DSPs.
Now, Friedman said, the addition of machine learning allows the flow of ad requests to DSPs to be modified in real time, based on each one’s ongoing record of bidding behavior. Smaato claims that its ATC can now reduce undesirable bids by as much as 72 percent, because of this better DSP targeting.
This allows the decisions about which inventory goes to which DSPs to be continually modified on the fly, instead of following static rules.
He added that, while he assumes other exchanges are similarly directing ad requests to specific DSPs, he isn’t aware of announcements by others indicating that they employ machine learning.
Why this matters to marketers. Guiding inventory ad requests to the most favorable bidders is the key goal for digital publishers, so Smaato’s enhanced ATC — if it works as advertised — could help publishers sell the most in-app inventory at the best prices. Fewer, more valuable inventory opportunities can mean higher bids, and could help publishers place all of their inventory.
Friedman says that the ATC intelligent funneling of ad requests also reduces time and expense for the DSPs, a savings that could then be passed on to the advertisers. He also says that the new system helps advertisers by matching their bids with the most relevant inventory.
Opinions expressed in this article are those of the guest author and not necessarily MarTech Today. Staff authors are listed here.